The H.264/AVC video coding standard offers several kinds of Intra prediction to reduce spatial redundancy between spatially adjacent blocks. In Intra prediction, pixels from already coded neighbouring blocks of a current block are used to predict the current block. For coding of the luminance signal, H.264 allows a total of nine Intra prediction modes for 4×4 and 8×8 blocks, and a total of four Intra prediction modes for 16×16 blocks. Similarly, a total of four Intra modes are usable to predict 8×8 chroma blocks for chrominance signals.
In recent years, several new Intra coding methods were proposed. For example in “Intra Prediction by Template Matching” by T. K. Tan, C. S, Boon and Y. Suzuki in Proc. ICIP'06, October 2006, template matching (TM) is used to search for matching blocks within the already coded areas of the current frame. In TM, both the encoder and the decoder perform the same motion search using already coded neighbouring pixels, therefore eliminating the need for coding the motion vectors. In “Extended Texture Prediction for H.264/AVC Intra Coding” by J. Balle and M. Wien in Proc. ICIP'07, October 2007, a combination of TM and Displaced Intra Prediction (DIP) are used. Similar to TM, DIP uses motion search to find matching blocks in the already coded areas of the current frame. Unlike TM, DIP performs motion search using the current block and sends the motion vectors to the decoder. Further, a new Intra coding method named Bidirectional Intra Prediction (BIP) was proposed by VCEG. BIP uses two techniques: combining two existing prediction modes to form bi-directional prediction modes, and allowing alternative coding order of 4×4 and 8×8 blocks within a macroblock (MB).
Conventional intra-prediction uses pixels from left and upper neighbour blocks to predict a MB. The MBs must be sequentially processed, since a reconstructed left and upper MB must be available for prediction. In at least all the above-mentioned Intra coding methods, the reconstructed pixels from the same slice as a current MB but not belonging to the current MB are necessary for Intra prediction of the current MB. Due to this characteristic, all the Intra MBs in a slice must be serially decoded.